Stroke-based Image Tonal Adjustment

Accompanied with the population of digital camera, the desire for tonal adjustment has been increasing significantly. However, existing local tonal adjustment tools such as Photoshop often suffer from the fact that the fine selection process of an object or a region is time-consuming and labor- intensive.This thesis proposes a stroke-based local tonal adjustment approach, the kernel of which is stroke-based edit propagation. It allows users to roughly paint on regions of interest and specify the edit parameter (such as color, brightness, saturation, etc.) as well as edit value on each region, and propagates these user constraints to the rest of the image without segmentation. Firstly, edit-value and edits-map are defined to generally support different image appearance edits. Then, an edge-preserving affinity for natural image with fuzzy object is presented to measure the similarity of pixel pairs. Guided by the principle that"similar pixels should receive similar edits", the approach builds an edit propagation energy formulation and solves it by minimizing the quadratic cost function. Finally, tonal parameters are adjusted in different color spaces with the guidance of edits-map.The innovation is that, by comparing region-based matting affinity and traditional global affinity, we design a new energy minimization formula to propagate edits in nature images with fuzzy objects. Moreover, we implement a local tonal adjustment tool in which parameters like color, brightness, saturation, etc. are adjusted in two device-independent color spaces (Lab & HSI) and comparison experiments between our method and previous ones are carried out. Judging from the experimental results, our method is an efficient method for local tonal adjustment in nature image.